Comparison
ChatGPT vs Claude for work | AI Course London
Choose between ChatGPT and Claude by testing the same permitted task in each. Use the same source, instructions and checking criteria, then compare factual accuracy, instruction following, usable output and review effort. Check the features, limits and data settings of the actual plans available to you. A general winner tells you less than a repeatable test on work you understand.
AI Course London · First published · Updated
What should a fair comparison hold constant?
Keep the task, source material, requested format and success criteria the same. Start a fresh conversation in each product so earlier messages do not supply one with extra context. Record the date, product, plan and model label shown, plus any tools used.
Compare like with like. An answer using live web search is a different experiment from an answer based only on supplied text. If a required feature is unavailable in one plan, record that practical limitation rather than pretending both attempts had identical conditions.
How can I test both tools on a realistic work task?
Use a short source with a few facts and one deliberately missing detail. Ask for an output that must preserve those boundaries. This makes it easier to spot a confident invention than an open-ended request for a “great” piece of writing.
Fictional customer-update comparison
Source material
Approved project facts: The sample pack was dispatched on Tuesday. The carrier reference is REF-104. The arrival date has not been confirmed. The customer asked whether Friday delivery is possible.
Prompt to try
Draft a customer update in at most 80 words using only these facts. Include the carrier reference. Do not promise an arrival date. End with one clear next step for our team. Then list any fact we must confirm before sending.
Human reference for checking
A usable draft says that the pack was dispatched on Tuesday and gives REF-104. It does not promise Friday delivery. It identifies the arrival date as something the team must check with the carrier before replying with a commitment.
This is a fictional practice exercise. The reference is written for checking your result; it is not an attendee outcome or a claimed model response.
What criteria should I use to score the answers?
Write your criteria before reading the results. Mark each one pass, needs correction or fail, with the actual phrase that caused your decision. Separate factual errors from a writing style you simply prefer.
| Criterion | Question to ask | Evidence to retain |
|---|---|---|
| Source accuracy | Does every factual claim match the source? | Unsupported or altered statement |
| Instruction following | Did it respect the format and exclusions? | Word count and missing constraints |
| Useful draft | Can a colleague act on this after review? | Required rewriting |
| Review effort | What did checking and correction take? | Measured minutes and edits |
| Practical fit | Is the required feature permitted and available? | Actual plan and workplace constraints |
How should I compare cost and privacy?
Use the current provider pages and your organisation’s approved setup. Look at the billing period, currency, tax, seat requirements and limits that matter for your tasks. Static prices in old articles are not a reliable basis for a purchase.
Data treatment can depend on the product, plan, settings and contractual arrangements. Do not infer permission to upload work material from a brand reputation or a broad claim that one provider is “private”. Ask the responsible person in your organisation when the approved route is unclear.
What should I do after the first test?
Repeat with one different task you actually perform, such as extracting actions rather than drafting an email. Keep the test small enough that you can check it thoroughly. Record variability instead of rerunning only the answer you dislike until you get a preferred result.
You may keep different tools for different jobs, or decide that the checking effort makes one task unsuitable for either. The useful decision is the setup you can use responsibly and consistently. Our workshop lets you practise those decisions with your own permitted examples.
Before you try it
Which tool is always better for writing?
There is no universal winner established by this guide. Compare drafts against your source, audience and voice. Record the revisions you need rather than relying on a general ranking.
Should I compare every available model?
Start with the products and plans you can actually use. A small, repeatable test with clear checks is more useful than trying every model without a decision criterion.
Can I use the fictional reference as a model benchmark?
No. It is a human-written checking aid for one exercise. It does not report a measured comparison of providers or predict how they will answer.
Check the source
Product features, plans and settings change. Check the provider and your organisation before using work material.
Provider information
Check the actual plan before choosing
Claude is the starting point in our public workshop. Check the current plan for your task.
Compare the product and plan with your required workflow and workplace permissions.
Check the current Notion AI features and plan terms if your team already uses Notion.
Claude and ChatGPT links go directly to the providers. The Notion link is an affiliate link; we may receive a commission if you make an eligible purchase.
Practise on your own work problem
Join Andrew for a one-day London workshop for non-technical professionals. Work in a group of at most eight, with guided practice from 9am to 4pm. The £599 fee includes lunch, materials and 30 days of support.
See workshop dates and details